Re: [Discuss-gnuradio] RF Gain and BB gain of XCVR2450

[Marcus D. Leech]

On 11/01/2010 04:19 PM, Jason Abele wrote:
>
> For the GNURadio drivers for either USRP2 or USRP1, the gain on
> receive is add in a loosely noise-figure optimized fashion by adding
> gain to the element closest to the antenna first and then filling in
> from there.  (The mathworks UDP driver should work like the GNURadio
> libusrp2 version)
>   
I think that if you really care about noise figure, though, you *have*
to use an
  external LNA.  I think at its *very best*, the XCVR2450 has a noise figure
  of about 8dB.  A GaAs HEMT or HJ-FET LNA in front of that with a
  sub-1dB noise figure and 15-20dB of gain would give you a much nicer
  Tsys than with the XCVR alone.


It's typically the case that RF VGAs use a fixed-gain amplifier
  followed by a switchable attenuator matrix.  (or worse, an
  attenuator matrix followed by a fixed-gain amplifier--yuck!).

For any serious weak-signal work, you pretty much have to use an
external LNA right
  out at the antenna.  None of the downconverter chips in common use for SDR
  have spectacular noise figures, because they're generally designed for
applications
  where noise figure isn't that important.  The XCVR2450 downconverter
(MAX2829), for example,
  was designed for WiFi applications, which don't have particularly
stringent noise figure requirements.

Similarly the DBS_RX MAX2118 (and now MAX2112) are designed for
direct-broadcast-satellite applications
  where there's a low-noise block-downconverter at the dish, and that
down-converter typically has
  a sub 0.8dB noise figure, and usually 40-50dB of gain.  The noise
figure of the direct-conversion part
  essentially vanishes when divided by the gain of the LNB, so there's
no attempt made by the designers of
  these chips to optimize noise figure to any great extent.

Any weak-signal work will *necessarily* require that the first low-noise
gain element be right up at the receive
  antenna, usually hung directly off the feed, with as little
feed-line/connectors as possible between the feed and
  the first gain stage.  That pretty-much means an external LNA,
although you *could* somehow mount a USRP2
  right up at the antenna, I'd be concerned about radiated "goop" from
the USRP2/USRP1 in that kind of scenario.
  And you'd still need a decent LNA in front of the
USRP2+daughtercard--none of them are particularly low noise.

-- 

Marcus Leech
Principal Investigator
Shirleys Bay Radio Astronomy Consortium
http://www.sbrac.org